By targeting nearby M dwarfs, a transit search using modest equipment is capable of discovering planets as small as 2 Earth radii in the habitable zones of their host stars. The MEarth Project, a future transit search, aims to employ a network of ground-based robotic telescopes to monitor M dwarfs in the northern hemisphere with sufficient precision and cadence to detect such planets. Here we investigate the design requirements for the MEarth Project. We evaluate the optimal bandpass, and the necessary field of view, telescope aperture, and telescope time allocation on a star-by-star basis, as is possible for the well-characterized nearby M dwarfs. Through these considerations, 1,976 late M dwarfs (R < 0.33 Rsun) emerge as favorable targets for transit monitoring. Based on an observational cadence and on total telescope time allocation tailored to recover 90% of transit signals from planets in habitable zone orbits, we find that a network of ten 30 cm telescopes could survey these 1,976 M dwarfs in less than 3 years. A null result from this survey would set an upper limit (at 99% confidence) of 17% for the rate of occurrence of planets larger than 2 Earth radii in the habitable zones of late M dwarfs, and even stronger constraints for planets lying closer than the habitable zone. If the true occurrence rate of habitable planets is 10%, the expected yield would be 2.6 planets.